5G, AI and IoT all place new demands on data center infrastructure – and there’s no room to cool off. While edge computing helps power these technologies, it also demands denser infrastructure in increasingly confined spaces. Boards, components and racks consume more power in tight spaces, leading to increased heat intensity that requires new cooling solutions.
The challenge is threefold: how will you deliver more computing power with less space while regulating thermal parameters for optimal performance?
New innovations and their resulting technology applications require new, and varied, approaches to thermal design. While originally limited to mainframe computers, liquid cooling has since made its way into the data center. Increasingly, companies consider liquid cooling to be a competitive differentiator.
It’s time to think about your data center roadmap and how you will adjust to greater computing demands and less operational space. As a leader in data center infrastructure, we’re helping companies implement data center solutions that provide low latency and complete coverage while keeping costs down. Liquid cooling will contribute to these goals, allowing high-density circuit designs that fully utilize the most powerful processors.
Why liquid cooling beats traditional air-cooling methods
To understand liquid cooling’s advantages, it’s important to consider why air cooling won't be enough in new data center environments. First, the next generation of CPUs and circuitry simply runs hotter. New data centers confined in smaller spaces put a premium on rack space and minimize airflow. In addition, an emphasis on energy efficiency demands new solutions. According to Global Market Insights, up to 40% of data center operating costs are due to HVAC systems.
Until now, the most common alternative to air cooling has been a liquid cooling technique using a combination of water and additives. However, as chemicals get added, the mixture becomes hazardous and needs to be treated as such during installation and disposal. Erosion and corrosion become common and can be particularly hard to spot or fix in tight spaces.
Plus, water temperature can vary as much as 10° in a chip-level, cold plate-based cooling system. This occurs because heat transfers from the chip raise the water temperature as it moves from one cold plate to the next. With new, high-density data centers, water-based cooling systems aren’t sustainable. It’s time for a solution that offers improved efficiency and consistency.
We’re in the era of refrigerant-based liquid cooling
Refrigerants are the future of liquid cooling for data center infrastructure. For starters, refrigerant pumps use 10% of the power of water pumps. Refrigerant maintains a significantly cooler temperature as it travels, with only a 2° variation in different parts of the rack, compared to the 10° variation of the water and additive mixture. The liquid also needs to be changed less frequently, lowering total cost of ownership and reducing time and labor spent on proper disposal.
There are several options when using refrigerant-based liquid cooling in a data center. For example, Data Center Frontier explains how Google incorporated touch-cooling, also known as chip level cooling, into its data centers that perform AI data crunching last year. Touch-cooling uses a cold plate that comes in contact with the chips. Other forms of liquid cooling include back of rack cooling, which works like an automotive radiator or refrigerator. This dual-phase cooling uses refrigerant to cool racks and chips. Immersion cooling, an emerging trend, completely immerses the rack or server using dielectric liquid.
Custom cooling makes your data center future-ready
We can provide liquid cooling solutions tailored to your specific infrastructure, workloads and data center environments. We design, develop and build products with the highest standards for reliability. Our designers and engineers integrate liquid cooling to make sure components perform at their peak, helping you avoid unnecessary throttling while minimizing temperature fluctuation and extending component life.
With our global capabilities, we can help you deliver a full scope of data center offerings from Sketch to Scale® — starting with designs that optimize rack/server density. It’s a one-stop approach that delivers the highest quality.